Population Ecology of the Eastern Wild Turkey in Northern Missouri Author(s): Larry D. Vangilder and Eric W. Kurzejeski Source: Wildlife Monographs , Oct., 1995 , No. 130, Population Ecology of the Eastern Wild Turkey in Northern Missouri (Oct., 1995), pp. 3-50 Published by: Wiley on behalf of the Wildlife Society Stable URL: https://www.jstor.org/stable/3830761 REFERENCES Linked references are available on JSTOR for this article: https://www.jstor.org/stable/3830761?seq=1&cid=pdf- reference#references_tab_contents You may need to log in to JSTOR to access the linked references. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org. Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at https://about.jstor.org/terms Wiley and Wildlife Society are collaborating with JSTOR to digitize, preserve and extend access to Wildlife Monographs This content downloaded from 128.227.171.90 on Wed, 08 May 2024 22:10:33 +00:00 All use subject to https://about.jstor.org/terms POPULATION ECOLOGY OF THE EASTERN WILD TURKEY IN NORTHERN MISSOURI by LARRY D. VANGILDER AND ERIC W. KURZEJESKI NO. 130 OCTOBER 1995 WILDLIFE MONOGRAPH (ISSN:0084-01 73) A Publication of The Wildlife Society This content downloaded from 128.227.171.90 on Wed, 08 May 2024 22:10:33 +00:00 All use subject to https://about.jstor.org/terms t : st 0#e v FRONMSPIECE. Adult eastem wild turkey gobblers. Population modeling suggests that varying levels of spring gobbler harvest affect the age structure of the male segment of the population. The level of spring gobbler harvest, therefore, has qualitative implications in harvest management decisions. (Photo by Jim Rathert, Missouri Department of Conservation.) L: F4 :s This content downloaded from 128.227.171.90 on Wed, 08 May 2024 22:10:33 +00:00 All use subject to https://about.jstor.org/terms POPULATION ECOLOGY OF THE EASTERN WILD TURKEY IN NORTHERN MISSOURI LARRY D. VANGILDER Missouri Department of Conservation, Fish and Wildlife Research Center, 1110 South College Avenue, Columbia, MO 65201 ERIC W. KURZEJESKI Missouri Department of Conservation, Fish and Wildlife Research Center, 1110 South College Avenue, Columbia, MO 65201 Abstract: The success of eastern wild turkey (Meleagris gallopavo silvestris) restoration programs in the Midwest has led to dramatic increases in occupied range. Habitats with limited forest land, once thought unsuitable for wild turkey, now support densities that exceed 30 birds/km2 of timber. Turkey population growth has resulted in the liberalization of harvest regulations with corresponding increases in hunter numbers and turkey harvest. As pressures on the wild turkey resource escalate, so does the need for an enhanced understanding of population dynamics. We used radiotelemetry to investigate survival and reproduction of wild turkey hens from 1980 to 1989 in the mixed timber-agricultural region of north-central Missouri. We also used direct recovery rates of banded gobblers to examine spring harvest mortality. Radio-marked hens (n = 327) were monitored annually, -5 times/week, to derive estimates of cause-specific mortality, seasonal survival, and reproductive parameters. Band numbers from tagged gobblers (n = 134) were recorded at mandatory turkey check stations to estimate harvest mortality. Seasonal survival rates of hens were variable, with winter rates exhibiting the greatest range (0.625-1.00). Although daily survival rates did not differ among seasons (P = 0.142) according to analysis of variance, a model selection procedure indicated that seasons could not be combined. Cause-specific mortality varied among seasons within years, with predation being the major cause of mortality. Legal fall-harvest mortality averaged 4.4% of radio-marked hens. Illegal kill of hens during the spring gobbler season ranged from 0.0 to 30.0% and averaged 5.2%, exceeding legal fall harvest of hens in some years. During the first year after banding, 17.2% of adult gobblers were harvested during spring hunting season. The nesting rate (measured as the proportion of hens attempting to nest) was consistent and high (>0.9) across years; however, hen success and nest success showed significant annual variation. Hen success differed between adult and subadult hens as did the success of first nests. We observed no difference in the success of renests between age classes of hens. Most poult mortality occurred by 2-weeks posthatch; however, annual estimates varied among years. Poult mortality at 4-weeks posthatch ranged from 0.419 to 0.705. Across years, increasing poult mortality was associated (p = 0.75, P = 0.054) with the number of days in June that rainfall exceeded 2.54 cm. We used a stochastic population model to examine the effects of simulated spring harvest on the age structure of males and the effects of fall harvest on population growth. Increasing levels of spring harvest resulted in a decreasing proportion of adult males in the spring population and in the harvest. At a 5% level of fall harvest, population size at year 40 increased in 72% of the simulations. When a 10% level of fall harvest was modeled, population size at year 40 decreased in all simulations. The model was sensitive to changes in hen survival, nest success, and poult mortality. Simulations were run for 4() years with an initial population size of 5,820. Decreases in poult mortality of 10 and 20% resulted in increases in average simulated population size at year 40 of 3,154 and 19,957% (32.54 and 200.57 fold), respectively. The variation we observed in wild turkey survival and reproduction and our population model outputs suggest that harvest strategies can affect population growth and age structure. Fall harvests have the greatest potential to affect population growth. We recommend that state-level estimates of harvest, hunting pressure, reproduction, and survival form the basis for developing and evaluating wild turkey harvest strategies in the Midwest. A conservative approach to spring and fall harvest regulations is recommended if sufficient data to model populations are not available. WILDL. MONOGR. 130, 1-50 Key words: cause-specific mortality, eastern wild turkey, fall harvest, hen success, Meleagris gallopavo silvestris, northern Missouri, population modeling, poult mortality, seasonal survival, spring harvest. This content downloaded from 128.227.171.90 on Wed, 08 May 2024 22:10:33 +00:00 All use subject to https://about.jstor.org/terms > ) WILDLIFE MONOGRAPHS CONTENTS INTRODUCTION 6 Acknowledgments 7 STUDY AREA 7 METHODS 9 Survival Estimation 11 Nesting 1 2 Harvest and Hunting Pressure 13 Additional Statistical Analyses 13 Population Model 14 Simutated Sp7 ing fIarvest 14 Simulated Fall Harvest 15 Sensitivity Analyses 16 RESULTS 1 7 Annual Survival 17 Causes of Death 17 Seasonal Survival and Cause-specific Mortaiity 1 7 Model Selection for Survival Data 19 Legal and Illegal Hen Harvest 21 Harvest of Adutt Gobblers During Spring Hunting Season 23 Chronology of Nesting 23 Nesting Rate 23 Hen Success 23 Nest Success 24 Clutch Size 25 Hatchability and Poult Mortality 27 Body Weights 27 Population Modeling 28 Sim2llated Spring Harvest 28 Simulated Fall Harvest 29 Sensitivaty Analyses 29 DISCUSSION 30 Annual Survival 30 Seasonal Survival 31 Cause-specific Mortality 32 Spring Gobbler Harvest Mortality 33 Fall Harvest Mortality 33 Nesting Chronology 34 Nesting Rate 35 Hen Success 35 Nest Sucoess 35 Clutch Size, Hatchability, and Poult Mortality 36 Body Weight 36 Population Model 37 Ssmtllated Spring Harvest 37 SimuSted Fall Harvest 38 Sensitivity Anslyses 39 MANAGEMENT RECOMMENDATIONS 39 LiTERATURE CITED 41 APPENDIXES 43 INTRODUCTION Eastern wild turkey (Meleagris gallo- pavo silvestris) restoration in the Midwest using wild-trapped birds, began in the mid- 1950's and early 1960's (Dickneite 1973, Donohoe and McKibben 197S, Little 1980). Initially, restoration efforts were concen- trated in heavily forested areas because it was thought that turkeys required exten- sive tracts of forest land, particularly for fall and winter foods (Ellis and Lewis 1967). In the 1960>s, several experimental turkey releases were made in areas that were <50% forested, amid skepticism that wild turkeys could not develop self-sustaining populations in these habitats (Little 1980 McMahon and Johnson 1980)* The success of these early efforts led to the conclusion that the eastern wild turkey could survive under a greater variety of habitat condi- tions than had previQusly been thought (Klonglan et al. 1970, Little 1980). Em- phasis on restoration within the Midwest shifted from heavily timbered areas to those with a greater mixture of timber and agricultural lands (Little 1980> Little and Varland 1981> Miller et al. 1985). Subse- quent studies of habitat use in mixed tim- ber-agricultural cover types have shown crop lands provide an accessible and abun- dant winter food source (Crim 1981> Kur- zejeski and Lewis 1990). It has been sug- gested that access to row crops in winter improves both survival (Porter 1978 Por- ter et al. 198Q, Kurzejeski et al. 1987) and reproduction (Porter 1978, Porter et al. 1983). Densities of eastern wild turkeys in mixed timber-agricultural habitats of the Midwest have been estimated to exceed 30 birds/km2 of timber (Hanson 1984? Kur- zejeski et al. 1987, Little et al. 1990) com- pared to densities that are one-third to one- half as high in the heavily forested portions of this region (Natl. Wild Turkey Fed. 1986). The demand for consumptive use and the economic value of wild turkeys has This content downloaded from 128.227.171.90 on Wed, 08 May 2024 22:10:33 +00:00 All use subject to https://about.jstor.org/terms WILD TURKEY ECOLOGY Vangilder and Kurzejeski 7 increased dramatically as turkey popula- tions and occupied range have expanded. In Missouri, 90,624 turkey permit buyers spent over 10 million dollars in association with the 14-day 1988-spring turkey season (Baumann et al. 1990). Turkey hunter den- sities during the 1988 spring season in northern Missouri were estimated at 5.2 hunters/km2 of forest, a 1.6-fold increase from 1979 (L. D. Vangilder, unpubl. data). In addition, within the last decade, most Midwestern states have either instituted or liberalized fall either-sex turkey seasons. These factors have led to an increase in harvest pressure resulting in greater em- phasis on turkey population management. However, limited data on wild turkey pop- ulation parameters are available to support management decisions. Most turkey re- search in the Midwest has been conducted on expanding (Porter 1978, Porter et al. 1980, Kulowiec and Haufler 1985) or re- cently reintroduced populations (McMa- hon and Johnson 1980, Clark 1985, Miller et al. 1985). Within established popula- tions, Kurzejeski et al. (1987) presented 1-year's survival data and Vangilder et al. (1987) examined reproductive parameters over 4 years. Long-term population data are needed to develop a basis for sound population management decisions. The purpose of this paper is to expand on the work of Kurzejeski et al. (1987) and Vangilder et al. (1987) by examining 7 years of radiotelemetry data on wild tur- key hen survival and reproduction and 7 years of direct recovery rate data for adult gobblers from the spring wild turkey sea- son. Our objectives were to determine cause-specific and seasonal mortality rates, nesting rates, hen success, and poult mor- tality and to model the potential impacts of various spring and fall harvest strategies on population size and age structure. We also examined the relationship between spring weather and reproductive effort. Acknowledgments.-We thank the nu- merous individuals who assisted with field work, particularly turkey trapping. We ac- knowledge the support of V. L. Kimmel- Truitt, D. W. Murphy, B. G. Root, E. A. Keyser, J. S. Fleming, B. J. Otten, and all the personnel of the Missouri Department of Conservation, Kirksville, Missouri. T. G. Kulowiec was responsible for program- ming of the wild turkey computer simu- lation model. S. L. Sheriff provided input into study design and assisted with data analyses. K. R. Mitchell prepared the final drafts of the manuscript and graciously tolerated our frequent modifications. Many private landowners unselfishly permitted our around-the-clock trespass- ing over the numerous years of the study. Their support was essential to the project's success. Landowners E. R. Jayne, C. Tur- ner, C. J. Scriven, M. E. Wade, and C. L. Walters deserve special mention. C. J. Scri- ven frequently permitted use of his resi- dence and outbuildings to process cap- tured turkeys. M. E. Wade provided every imaginable form of support from use of his facilities, to dislodging our vehicles from field roads, to an occasional warm break- fast; his assistance was invaluable. The Missouri Department of Natural Resources granted access to some of their lands for trapping and provided assistance with handling and marking turkeys. The Na- tional Wild Turkey Federation (NWTF), the Missouri State Chapter of the NWTF, and numerous local chapters provided fi- nancial support and radiotelemetry equip- ment. G. W. Clark, president of the Mis- souri State Chapter of the NWTF during a significant portion of this study, provided major assistance in securing NWTF chap- ter funding. J. B. Lewis, who served as our supervisor during the project, deserves thanks for the countless hours of discussion that formed the basis for the success of this project. His vision enabled us to grasp the need for long-term research in wild turkey population dynamics, and his support en- abled the project to become a reality. W. F. Porter and D. E. Steffen critically re- viewed the manuscript and provided many useful comments and suggestions. STUDYAREA The study was conducted in a 52-km2 area approximately 8 km west of Kirksville in Adair County, Missouri (Fig. 1). Total This content downloaded from 128.227.171.90 on Wed, 08 May 2024 22:10:33 +00:00 All use subject to https://about.jstor.org/terms 8 WILDLIFE MONOGRAPHS g WOODLAND \ CH ARITON RIVER g GRASSL AND SEMIOPEN- WOODLAND SEMIOPEN- GRASSL AND CROP 5 km Fig. 1. Cover types on the 52-knr2 Adair County, Missouri, study area in 1983. land area in Adair County is 1,474 km2 of which forest composes 263.3 km2. In 1980, Adair County was occupied by 24,870 peo- ple, 17,167 of whom lived in the city of Kirksville (U.S. Dep. of Commerce Bur. of the Census 1981). Over 95S of the study area was in private ownership. The own- ership pattern was characterized by small (<80 ha) tracts with only 1 ownership ex- ceeding 400 ha. Most land was held for This content downloaded from 128.227.171.90 on Wed, 08 May 2024 22:10:33 +00:00 All use subject to https://about.jstor.org/terms WILD TURKEY ECOLOGY Vangilder and Kurzejeski 9 the production of income from agricul- tural crops. A few tracts were owned ex- clusively for recreational purposes, pri- marily hunting. Habitats were divided into 5 cover types (Fig. 1). Forest land composed 46% of the area and was dominated by large pole and small sawlog (>10.2 and <40.6 cm dbh) stands of oak (Quercus spp.) and hickory (Carya spp.). Forest distribution (Giess- man et al. 1986) was typical of areas in northern Missouri that support high turkey densities. Grazing occurred annually on approximately 255E of the forest lands. Pas- ture and hay land occurred on 24% of the study area. Tall fescue (Festuca spp.) was the dominant cool-season grass with or- chard grass (Dactylisglomerata), common timothy (Phleum pratense), and clovers (Trifolium spp.) occurring in < 15% of fields. Haying usually occurred twice be- tween 15 May and 31 July. Native warm- season-grass pasture, predominantly big bluestem (Andropogon gerardi), was pres- ent on approximately 54 ha. Crop lands composed 16S of the study area and were located primarily along the Chariton River (Fig. 1). Corn and soybeans were the dominant crops. Field size av- eraged 10.5 ha, and fall plowing was com- mon with >50% of fields tilled after har- vest. Open lands not used for agricultural purposes (dominated by grasses and hav- ing <25S woody cover) were classified as semiopen grassland and composed 9.9% of the area. Semiopen woodland, which oc- curred on 3.7S of the study area, included nonagricultural lands with >25% woody cover and an understory of herbaceous plants. Common woody species in the se- miopen habitats were Indiancurrant cor- alberry (Symp7ioricarpos orbiculatus), shingle oak (Quercus imbricaria), and black locust (Robinia pseudoacacia). Topography on the study area was roll- ing with elevations ranging from 226 to 297 m. Annual precipitation at Kirksville ranged from 8.8 to 19.7 cm (1980-88). Mean minimum temperatures ranged from -14.6 to -5.7 C in January and -9.6 to -2.1 C in February. Annual snowfall ranged from 7 6 to 70.1 cm. Mean maxi- mum summer temperatures ranged from 27.6 to 34.4 C in July and from 25.9 to 3t3.9 C in August. Wild turkeys were first released in Adair County in December 1961 and January 1962. These releases consisted of 6 adult and s subadult gobblers, and 5 adult and 8 subadult hens. The birds were captured at 4 different sites in the Missouri Ozarks. All birds were released at 1 site on private land located within our study area. In 1967 the first spring gobbler season was held in Adair County with a harvest of 35 birds (Table 1). Fall hunting was not instituted until 1979. Since 1979, both spring and fall turkey seasons have been 14 days in length. Throughout the study, the spring gobbler season opened on the Monday closest to 21 April, whereas the fall hunting season oc- curred during a 2-week period ending the last Sunday in October (except in 1981 when it closed on 1 Nov). The spring hunt- ing season had a 2-gobbler bag limit and restricted hunters to 1 birdXweek. The fall season bag limit was 1 bird of either sex from 1980 to 1985 and 2 birds (with a 1 bird/week provision) beginning in 1986. METHODS Beginning in the summer of 1980, tur- keys were trapped during 2 periods: sum- mer (Aug-Oct) and winter (Dec-Mar). Trapping occurred annually from 1980 to 1988 with the exception of April 1982 through July 1983. Both cannon and rocket nets were used to capture turkeys. After its sex and age were determined each bird was marked and released at the capture site. During the summer trapping period hens were classified as poults or adults. Sex of poults was determined by cloacal ex- amination. During the winter trapping pe- riod, hens were classified as subadults (first winter of life) or adults (second or later winter of life). Age was determined by standard methods (Pelham and Dickson 1992:40). Thus, subadult hens were 6-10 months old at capture. Birds were weighed to the nearest 0.1 kg, except during the winter trapping season of 1984-85. Males were marked with a 2.5-cm, round, cattle This content downloaded from 128.227.171.90 on Wed, 08 May 2024 22:10:33 +00:00 All use subject to https://about.jstor.org/terms Table 1. Total turkey harvest, turkey harvest per square kilometer of forest, number of hunters, and number of hunter trips for spri County, Missouri, 1967-88. Spring season Fall firearms se Harvest/km2 No. hunters/km2 Harvest/km2 No. Year Harvest forest No. hunters forest No. trips Harvest forest No. hu 1967a 35 0.13 227 0.86 1968 34 0.13 343 1.30 1969 39 0.15 270 1.03 1970 70 0.27 380 1.44 1971 1 12 0.42 498 1.89 1972 141 0.54 625 2.38 1973 173 0.66 549 2.09 1974 150 0.57 594 2.26 1975 175 0.67 663 2.52 1976 241 0.92 614 2.33 1977 282 1.07 828 3.15 1978 286 1.09 762 2.90 1979b 413 1.57 877 3.33 394 1.5 1980 416 1.58 736 2.80 398 1.5 1981 582 2.21 907 3.45 366 1.4 1982 517 1.97 957 3.64 351 1.3 668 1983 436 1.66 432 1.6 1984 378 1.44 1,251 4.80 6,268 460 1.7 1985 535 2.03 567 2.2 1986 818 3.11 1,262 4.80 5>546 945 3.6 1)20 1987 894 3.40 1,025 3.9 996 3 1988 644 2.45 1)367 5.20 5,087 920 3.5 a First year of spring firearms turkey hunting in Adair County. b First year of fall firearms turkey hunting in Adair County. This content downloaded from 128.227.171.90 on Wed, 08 May 2024 22:10:33 +00:00 All use subject to https://about.jstor.org/terms WILD TURKEY ECOLOGY-Vangilder and Kurzejeski ll ear tag and individually numbered alumi- num bands on the patagium of both wings. Hens were marked with 2-aluminum pa- tagial bands and instrumented with back- pack-style transmitters. Poults were not ra- diomarked. During 1980-82, transmitters were a nonmortality-mode radio package weighing 130 g. During 1984-88, a 140-g mortality-mode transmitter was used. Survivai Estimation We monitored survival of instrumented hens >5 times/week using hand-held and vehicle-mounted antennae and portable scanning receivers. When a mortality sig- nal was received, cause of death was in- vestigated within 12 hours. When non- mortality transmitters were in use, daily locations of each hen were determined us- ing the intersection of 2 compass bearings from fixed telemetry stations. After 2-consecutive relocations indicated no movement, the status of the bird was in- vestigated to determine if a mortality had occurred. In these instances, cause of death was determined within 24-36 hours. Ev- idence at the transmitter recovery site was used to identify cause of death. If the car- cass was recovered intact, a necropsy was performed. During the nesting season, hens that did not move (nonmortality-mode transmitters) or hens with transmitters emitting a mortality signal were not dis- turbed if they were thought to be nesting. If, after 28 days, the hen had still not moved, cause of death was then investi- gated. We calculated estimates of annual sur- vival and survival distributions with the Kaplan-Meier staggered entry approach (Pollock et al. 1989) using program STAG- KAM (T. G. Kulowiec, Mo. Dep. Conserv., Columbia). The interval for which birds were considered at risk was 1 day. The variance of annual survival was calculated using Greenwood's formula (Pollock et al. 1989:9; equation [3]). l:)ifferences between adult and subadult survival distributions generated by the Kaplan-Meier approach were tested using the log-rank test (Pollock et al. 1989). We used a SAS program (L. W. Burger, Jr., The School of Nat. Resour., Univ. of Mo., Columbia), which uses files generated by STAGKAM, to calculate summary statistics for the log-rank test and the 3 chi-square tests. In a separate anal- ysis, the relationship of body weight at capture to survival time was tested on a subset of hens with a Cox proportional haz- ards model (Cox 1972, Cox and Oakes 1984: 91) using a microcomputer program mod- ified from Klotz and Meyer (1985:55-74). In this analysis, data from all years were combined and entry of individuals into the analysis was not staggered. We estimated annual,, seasonal, and cause-specific rates of mortality using pro- gram MICROMORT (Heisey and Fuller 1985), which is based on the Mayfield method (Mayfield 1961, 1975; Johnson 1979). Data also were entered into MI- CROMORT using staggered entry, and ra- dio-days for censored animals were in- cluded up until the time they disappeared as recommended by Vangilder and Sheriff (1990). Differences in annual survival rates between subadult and adult hens were test- ed by using log-likelihoods generated by MICROMORT for a full (age classes sep- arate) and reduced (age classes combined) model. Two times the absolute value of the differences in the log-likelihood val- ues for the 2 models provided a chi-square test statistic. Hens that did not survive >7 days postinstrumentation were deleted from the analyses. Seasonal intervals were defined as spring (15 Mar-31 May), summer (1 Jun-31 Aug) fall (1 Sep-30 Nov) and winter (1 Dec- 14 Mar) These intervals were chosen be- cause they corresponded with major bio- logical events in the allnual cycle of hens or with distinct shifts in habitat use (Kur- zejeski and Lewis 1990). Breakup of winter flocks occurred in mid- to late March fol- lowed by breeding, egg laying, and incu- bation. The summer period corresponded with the brood-rearing period. Changes in habitat use occurred among all seasons (Kurzejeski and Lewis 1990), but differed most among fall, winter, and spring. Use of pasture and old-field habitats declined This content downloaded from 128.227.171.90 on Wed, 08 May 2024 22:10:33 +00:00 All use subject to https://about.jstor.org/terms 12 tions were determined from 2-intersecting compass bearings taken at fixed stations. Highly localized movements were as- sumed to be indicative of egg laying. A hen that exhibited a highly localized pat- tern of movements was considered to have made a nesting attempt even if the hen never reached incubation. The nesting rate, then, was the proportion of hens that we considered to have made a nesting at- tempt. When no movement occurred for 2-consecutive days, incubation was as- sumed and the relative nest location was determined. To minimize disturbance of a hen, a circle <50 m in diameter was flagged around the nest. We continued to monitor daily locations of the hen until movement away from the nest of > 8 hours was noted. The nest was then located and the fate and clutch size determined. Un- successful nests were classified as lost either to predation or abandonment. Renesting attempts were monitored using the same methods. The number of poults hatched was determined from examination of egg- shells at the nest site. We defined hatch- ability as the number of eggs in successful nests that hatched. Poult mortality was measured by flushing each hen and its brood at 2- and 4-weeks posthatch. Poult mortality was not monitored after 4-weeks posthatch due to the tendency of hens to form brood flocks. Estimates of poult mor- tality did not include a hen and its brood if the hen was with other hens at the 2- or 4-week flush. We employed an approach developed by Cowardin et al. (1985) and used by Vangilder et al. (1987) to analyze nesting data. We tested for differences in hen suc- cess, first-nest and renest success, and re- nesting rate among years and between age classes. Only hens that attempted to nest and survived through the reproductive pe- riod (30 Jun) or hatched a clutch were included in these analyses. Hen success (H) is defined as the proportion of hens suc- cessful in 1 nesting attempt. Nest success is defined as the proportion of nests in which at least 1 egg hatched. Renesting rate is defined as the proportion of hens nesting a second time after losing the first nest. To avoid division by zero because no in the fall, and use of forest land increased. As the availability of oak mast declined, hens shifted from extensive use of forest to use of crop land. Distinct shifts in home ranges occurred in early winter as large flocks (>200) formed in association with crop lands. Extensive use of crop land was common until the flocks began to disperse in mid-March. Differences in seasonal sur- vival rates were tested by comparing daily rates of survival in each season in a 2-way analysis of variance (ANOVA) with year and season as main effects and the inter- action of year and season as the error term (general linear models procedures [SAS Inst., Inc. 1985]). We also used program MICROMORT to examine various models of our survival data. For examination of these models, age classes were combined. Initially, we con- sidered a model with 7 years of survival data (yrs were considered classes in pro- gram MICROMORT), 4 seasons (time in- tervals in program MICROMORT), and 5 causes of death (rate parameters in pro- gram MICROMORT). This model (140 parameters) was considered our global model because it had many parameters and a flexible structure (see Lebreton et al. 1992:83). Several models that were sub- sets of the global model were examined. The number of parameters was reduced for these models by pooling combinations of years, seasons, and/or causes. The most restrictive model, then, was a model with years, seasons, and causes of mortality combined (1 parameter was estimated). To determine the most parsimonious model for our survival data, we used Akaike's Information Criteria (AIC) as a basis for model selection (Lebreton et al. 1992:84). The AIC was defined as AIC =-2 InL + 2 x (no. of parameters) where L = the log-likelihood. For each program, MICROMORT provides a value for -2 lnL and for the number of param- eters estimated. Nesting We located each hen daily beginning 1 April to determine nesting status. Loca- This content downloaded from 128.227.171.90 on Wed, 08 May 2024 22:10:33 +00:00 All use subject to https://about.jstor.org/terms i=l WILD TURKEY ECOLOGY Vangilder and Kurzejeski 13 subadult hens nested successfully in 1982, data for these nesting parameters were transformed by H= (Ns + 05) (N+ 1) where N = the number of hens or nests, NS = the number of hens successful, the number of nests successful, or the number of hens renesting, and H = the transformed estimate of hen suc- cess, nest success, or renesting rate. Estimates were weighted by N . (H[1 - H]) because of variation in sample size among years and ages (Cowardin et al. 1985:24). We used General Linear Models procedures (SAS Inst, Inc. 1985) with year and age as main effects (2-way ANOVA) to test for differences. The interaction of year and age was used as the error term. We examined the relationship between various nesting-season parameters and weather data using rank correlation anal- yses (Spearman's rho) (Conover 1971:245- 249). Daily weather records were recorded at Kirksville, 8 km east of the study area. For all analyses, statistical significance is at the P < 0.10 level except in the case of multiple comparisons where P was set at 0.1 . m where m is the number of com- parisons made (Neter and Wasserman 1974:480). Harvest and Hunting Pressure Legal harvest mortality of hens and gob- blers was monitored through recovery of marked individuals at mandatory turkey check stations. Age ratios of both spring and fall harvest also were determined from check-station data. Hunters per square ki- lometer of forest were estimated from a state-wide, postseason, mail survey of fall turkey permit buyers in 1982, 1986, and 1987 and from spring turkey permit buy- ers in 1967-82, 1984, 1986, and 1988. We assumed that county-level information, derived from these surveys, approximated that within the study area for both age composition of the harvest and hunting pressure estimates. Both habitat and land ownership patterns in our study area were typical of those in portions of Adair Coun- ty that supported high-density turkey pop- ulations. Additional Statistical Analyses For parameters that were estimated us- ing proportions (e.g., hen success = no. hens successful . no. hens attempting to nest) within years, a ratio estimator was used to calculate an overall ratio estimate (across years) and its associated standard error. The formulas were as follows (Sne- decor and Cochran 1967:537): n : Ys i=l , __ n z Xi i=l and SE(r) / n n Nyi2+r2Nxi2 i=l i=l n - 2r : xiCi x2n(n-1) svhere r = ratio estimate, Yi = number of hens successful in year i, xt = number of hens attempting to nest in year i, x = mean number of hens attempting to n 1: XtX nest i=l , and n = number of years. The formulas were used to calculate ratio estimates and their standard errors across years for legal fall harvest, illegal kill of hens during spring gobbler season, direct recovery rates of adult gobblers, hen suc- cess, Srst-nest success, renest success, re- nesting rate, and renesting rate of hens disrupted during laying or incubation. For hatchability, poult mortality at 2 weeks, and poult mortality at 4 weeks, the ratio estimation technique was used to es- This content downloaded from 128.227.171.90 on Wed, 08 May 2024 22:10:33 +00:00 All use subject to https://about.jstor.org/terms 14 WILDLIFE MONOGRAPHS timate hatchability or survival within each year (in these cases, n = no. of clutches or broods; xi = no. of eggs hatching or poults dying at 2 or 4 weeks; Yt - no. of eggs in a clutch or no. of poults at hatching). Across year estimates and associated standard er- rors of hatchability, poult mortality at 2 weeks, and poult mortality at 4 weeks were calculated using a weighted mean (weight- ed by sample size). To determine whether there were among-year differences in hatchability or poult mortality at 2 or 4 weeks, we used the methods of Sauer and Williams (1989) and program CONTRAST (Hines and Sauer 1989). The program calculated an overall chi-square test for among-year dif- ferences. Contrast vectors were then used to make pair-wise (1 df) comparisons among years. The Bonferroni approach of using oe . m as the significance level was used to maintain the overall level of ex- periment-wise error (Hines and Sauer 1989: 3). In our case ol = 0.10 and m = 21 (m = no. of comparisons with n = 7 yrs of data). Thus the significance level for the pair- wise comparisons was 0.0048. Population Model A 2-age class (subadult adult) account- ing-style turkey-population dynamics model was developed to model our telem- etry data (T. G. Kulowiec and L. D. Van- gilder, Mo. Dep. Conserv., Columbia, un- publ. data). The model is both determin- istic and stochastic, in that it can be run for a user-specified number of data years for which the user provides input param- eters as well as for any number of years beyond the data years. Input for years be- yond the data years is randomly generated from a uniform distribution within 1 stan- dard deviation of the mean of survival and reproductive rates calculated from the data years. A uniform distribution was selected for use in this model rather than a normal distribution because we wanted parameter values to be randomly selected with equal probability from the range of values mea- sured during the data years. Use of the uniform distribution in the model allowed parameter values to vary more then they would have had we used a normal distri- bution. Thus, we considered use of the uni- form distribution to be conservative. The kinds of data required to run the model are presented in Table 2. Complete documentation for the model can be obtained from the authors. A brief description of how the model works can be found in Appendix A. The model pre- sented in Vangilder (1992:159-163)is sim- ilar to the model used in this study except that in Vangilder (1992:159-163) input for years beyond the data years was randomly generated from a uniform distribution bounded by the 95% confidence limits (in- stead of the standard deviation which was used in this study) around the means of survival and reproductive rates calculated for the data years. The model makes 3 interrelated as- sumptions (Vangilder 1992:159-161): 1. True parameter values vary indepen- dently (e.g., high annual survival is not associated with low hen success). 2. Hunting mortality is completely addi- tive to other forms of mortality. 03. Survival and reproduction are density- independent. Parameter estimates (survival and re- productive rates) for hens were derived from the 7 years of radiotelemetry data. Survival rate estimates for subadult and adult males for corresponding years were obtained from a study conducted in south- ern Iowa (1:). H. Jackson, Ia. Dep. Nat. Resour., Bocxne, unpubl. data). The values for these parameters and other model in- puts are listed in Appendixes B, C, and D. The natural mortality rates shown in Ap- pendix D are the daily rates of mortality for each season (as calculated by MICRO- MORT) for all causes of mortality except legal harvest. Simulated Spring Harvest. We used the model to examine the effects of sim- ulated spring harvest on the age structure of males. To do this, we varied the per- centage of males harvested from 0 to 60%, in 5% increments. The equations (Appen- dix A; eqs. 9-12) used to calculate spring This content downloaded from 128.227.171.90 on Wed, 08 May 2024 22:10:33 +00:00 All use subject to https://about.jstor.org/terms Table 2. Input parameters for a turkey population model used to examine the effects of simulated fall and spring harvest on population growth and age structure. Dates 1. Starting dates for each of 4 seasons (winter, spring, summer, and fall). 2. Starting and ending dates of hunting seasons (spring and/or fall). 3. The starting year of the simulation. 4. The starting date of the simulation. 5. The number of years for which input data will be supplied ("data years"). 6. The number of years beyond the data years that will be simulated by the model. Productivity 1. Age-specific (subadult, adult) nesting rates. 2. Age-specific first-nest success rate. 3. Age-specific clutch size-(first nests). 4. Age-specific hatchability (first nests). 5. Age-specific renesting rate. 6. Age-specific renest success rate. 7. Age-specific clutch size (renests). 8. Age-specific hatchability rate (renests). 9. Sex ratio of poults at hatching. Natural mortality 1. Poult mortality rates. 2. Age- and sex-specific daily mortality rates for each season. Hunting mortality 1. Number of birds shot or proportion of population shot (proportion of males for spring season; proportion of total population for fall season). Starting population size 1. Number of individuals in each age and sex class. WILD TURKEY ECOLOGY Vangilder and KurzeXeski 15 harvest for each age class (subadult, adult) incorporate the assumption that adult gob- blers are about 2 times more vulnerable to spring harvest than are subadult gobblers. This differential vulnerability produced model results that mimic the age structure of Missouri's spring harvest (L. D. Van- gilder, unpubl. data). In addition, Lewis (see Vangilder 1992:156) found, in a study in central Missouri, that direct recovery rates of adult gobblers were double that of subadult gobblers. At each level of spring harvest, we ran the model for 40 years. For the 33 years of simulated data, the same set of random numbers was used in each run to permit direct comparisons of age structure among levels of harvest. Be- cause we were primarily interested in changes in male age structure at a given level of hen survival and reproductive rates, only 1 run of the model at each level of spring harvest was necessary. For each lev- el of harvest, we calculated average per- cent of adult males in the prehunt spring population, average percent adult males in the spring harvest, average total harvest, and average annual survival rates (natural and natural + hunting) for adult and sub- adult males over the 40 years. We incor- porated a 5S level of fall harvest into all model runs. Simulated Fall Harvest.-To examine the impacts of fall harvest, we ran the model 25 times for each of 5 levels of fall harvest (0, 5, 10, 15, and 20%). The equa- tions (Appendix A; eqs. 1-8) used to cal- culate the fall harvest for each age class incorporate the assumption that subadult males, subadult females, and adult females are 2.33, 2.00, and 1.33 times, respectively, more vulnerable to fall harvest than are adult males. These differential vulnerabil- ities produced. model results that mimic the age structure in Missouri's fall harvests (L. D. Vangilder, unpubl. data). In the model, the proportion of suba- dults in the simulated